Download Seasonal hydrologic responses to climate change in the Pacific

PUBLICATIONS
Water Resources Research
RESEARCH ARTICLE
10.1002/2014WR015909
Key Points:
Hydrologic sensitivities characterize
seasons/locations susceptible to
future change
Transitional watersheds experience
greatest seasonal shifts in runoff
Seasonal streamflow estimation
works well when linearity and
superposition apply
Supporting Information:
Supporting Information S1
Correspondence to:
J. A. Vano,
[email protected]
Citation:
Vano, J. A., B. Nijssen, and D. P.
Lettenmaier (2015), Seasonal
hydrologic responses to climate
change in the Pacific Northwest, Water
Resour. Res., 51, 1959–1976,
doi:10.1002/2014WR015909.
Received 27 MAY 2014
Accepted 9 FEB 2015
Accepted article online 18 FEB 2015
Published online 3 APR 2015
Seasonal hydrologic responses to climate change in the Pacific
Northwest
Julie A. Vano1,2, Bart Nijssen1, and Dennis P. Lettenmaier1,3
1
Department of Civil and Environmental Engineering, University of Washington, Seattle, Washington, USA, 2Now at
College of Earth, Ocean, and Atmospheric Sciences, Oregon State University, Corvallis, Oregon, USA, 3Now at Department
of Geography, University of California Los Angeles, Los Angeles, California, USA
Abstract Increased temperatures and changes in precipitation will result in fundamental changes in the
seasonal distribution of streamflow in the Pacific Northwest and will have serious implications for water
resources management. To better understand local impacts of regional climate change, we conducted
model experiments to determine hydrologic sensitivities of annual, seasonal, and monthly runoff to
imposed annual and seasonal changes in precipitation and temperature. We used the Variable Infiltration
Capacity (VIC) land-surface hydrology model applied at 1/16 latitude-longitude spatial resolution over the
Pacific Northwest (PNW), a scale sufficient to support analyses at the hydrologic unit code eight (HUC-8)
basin level. These experiments resolve the spatial character of the sensitivity of future water supply to precipitation and temperature changes by identifying the seasons and locations where climate change will
have the biggest impact on runoff. The PNW exhibited a diversity of responses, where transitional (intermediate elevation) watersheds experience the greatest seasonal shifts in runoff in response to cool season
warming. We also developed a methodology that uses these hydrologic sensitivities as basin-specific transfer functions to estimate future changes in long-term mean monthly hydrographs directly from climate
model output of precipitation and temperature. When principles of linearity and superposition apply, these
transfer functions can provide feasible first-order estimates of the likely nature of future seasonal streamflow change without performing downscaling and detailed model simulations.
1. Introduction
The Pacific Northwest (PNW), defined as the Columbia River basin and its coastal drainages (Figure 1), has
strong seasonal variations in runoff. This is due to the combined effects of a Mediterranean climate, in which
most precipitation occurs in the fall and winter, and complex topography, wherein snow accumulates in
winter and melts in spring and summer. The region’s snowpack effectively serves as a natural reservoir that
slowly releases water throughout the dry season providing sustained streamflow critical for agriculture, fisheries, and municipal and industrial use. Manmade reservoirs also serve this purpose, but can hold only
30% of the region’s annual streamflow and depend on snowmelt for in-season resupply [Federal Columbia
River Power System (FCRPS), 2001].
A number of authors have previously studied the impact of climate variations on streamflow in the PNW
[e.g., Hamlet and Lettenmaier, 1999; Payne et al., 2004; Luce and Holden, 2009; Lee et al., 2009; Elsner et al.,
2010; Hamlet et al., 2010; Luce et al., 2013; Safeeq et al., 2014]. The motivations for these studies stem from
the region’s vast hydropower resources [Lee et al., 2009], dangers of flooding [Payne et al., 2004], endangered species issues related to anadromous fish [Mantua et al., 2010], agricultural water uses [Vano et al.,
2010], and the need to manage reservoir systems to balance these needs [Hamlet, 2011].
Many of these past studies have used output from Global Climate Models (GCMs), which typically have
native spatial resolutions on the order of 150–200 km, to force a hydrologic model offline. A key issue
faced by such studies is the need to translate inputs to a hydrology model, which operates at a higher
spatial resolution than the GCMs. In the PNW, as in other topographically complex regions, it is necessary to reflect the effects of topographic barriers such as the Cascade Mountain range that control local
climate. These issues have been addressed by downscaling techniques, most of which can also partly
account for the bias inherent in GCM simulations, and convert precipitation (P) and temperature (T)
VANO ET AL.
C 2015. American Geophysical Union. All Rights Reserved.
V
1959
Water Resources Research
10.1002/2014WR015909
Figure 1. Pacific Northwest region including the Columbia River basin and coastal drainages. Elevation (which ranges from sea level to
4400 m) is shown as grid cell averages at 1/16 resolution. Watershed units, described in section 3.2, are outlined in gray.
changes from GCMs to local P and T using statistics from observation-based data sets [e.g., Wood et al.,
2004; Abatzoglou and Brown, 2012].
These studies are, however, resource intensive and subject to uncertainties related to the subset of GCMs
used, the choice of hydrology model and the downscaling algorithm [Vano et al., 2014]. In addition, it is
challenging to determine whether differences associated with different GCM forcings are mostly due to
changes in P or T and to distinguish the season in which these changes have the largest impact on basinscale hydrologic predictions. This complicates the identification of regions that are more or less sensitive to
future change [Vano and Lettenmaier, 2014].
We address the challenges of identifying locations and seasons that are most vulnerable to future change
using an alternative, seasonal sensitivity-based, approach. We evaluate sensitivities to climate in terms of
precipitation elasticity (e), defined as the fractional change in runoff per fractional change in P, and temperature sensitivities (S), defined as the fractional change in runoff per C change in T. In essence, we perform
a sensitivity analysis of a hydrologic model, but instead of testing the sensitivity of model parameters, we
evaluate the modeled response of changes in forcing variables P and T.
Previous studies have also quantified hydrologic sensitivities [Schaake, 1990; Dooge, 1992; Dooge et al.,
1999; Sankarasubramanian et al., 2001; Fu et al., 2007; Vano et al., 2012, among others)]. Dooge [1992]
derived an elegant approach that formulated the elasticities with respect to annual runoff of P and potential
evapotranspiration. His approach is, however, complicated by the fact that potential evapotranspiration is
not readily measurable in practice and in models is a derived quantity that involves a number of dependent
variables. In contrast, surface air T is readily available for most GCMs, and is a metric most commonly
VANO ET AL.
C 2015. American Geophysical Union. All Rights Reserved.
V
1960
Water Resources Research
10.1002/2014WR015909
applied in climate change studies. Notably, temperature affects and is affected by hydrologic processes
such as evaporation and snow processes. However, as cautioned by Milly [1992], temperature is not a static
land-surface boundary condition, but rather is affected by latent heat exchange, something that is not
accounted for in the off-line simulations investigated here. Other approaches motivated by a more complete set of sensitivities (e.g., net longwave radiation, potential evapotranspiration) could perhaps be
explored, although we do not do so in this paper.
The seasonal sensitivity-based approach we outline is similar in concept to the approach used by Vano et al.
[2012] in the Colorado River basin, where modeled sensitivities were evaluated at an annual time scale. In
the PNW, however, a focus on the seasonal distribution of runoff and how it is impacted by seasonal
changes in P and T is particularly important. Despite abundant annual P over much of the PNW, P occurs in
the winter while water demand peaks in summer. The region’s aggregate reservoir storage is much smaller
than annual aggregate runoff, and reservoir storage is more distributed among multiple facilities than in
the Colorado River basin. As such, modest changes in the seasonal distribution of runoff have the potential
to affect reservoir system performance. Additionally, future climate projections in the PNW indicate that different seasons will experience greater changes than others [Abatzoglou et al., 2014; Mote et al., 2013] and
therefore understanding the impact of seasonal changes (e.g., cool versus warm season warming) is
valuable.
This study builds on the work of Elsner et al. [2010] and Das et al. [2011]. Elsner et al. [2010] classified PNW
watersheds into rain-dominated, transitional, and snow-dominated based on the ratio of peak snow water
equivalent (SWE) to accumulated winter (October to March) P. This classification scheme has been useful to
water managers in the region and therefore has been included in subsequent user-funded studies [e.g.,
Hamlet, 2011; Hamlet et al., 2013; Tohver et al., 2014]. In this work, we also create a classification scheme to
identify drivers of hydrologic change, but we base our analysis on modeled hydrologic sensitivities to P and
T changes rather than results derived from GCM output. Das et al. [2011] evaluated warm versus cool season
responses to T change in four major western U.S. river basins—the Colorado, Sacramento (North Sierra), San
Joaquin (South Sierra), and Columbia—using the Variable Infiltration Capacity (VIC) model applied at 1/8
spatial resolution. They found that among these four large river basins, three are more sensitive to warming
in the summer than in the winter, including the Columbia River. Das et al. [2011] did not, however, evaluate
changes in the seasonality of runoff, but rather focused on the sensitivity of annual runoff changes to warm
and cool season warming.
Our work is intended to provide a context for better understanding local-scale hydrologic change across
the PNW by applying modeled hydrologic sensitivity concepts to map spatial variations in e and S
across the region (section 4.1). We categorize hydrologic sensitivities at the HUC-8 unit scale (average
drainage area about 3000 km2) (section 4.2) and evaluate modeled monthly response to seasonal
changes in P and T (section 4.3). Finally, we evaluate the applicability of our monthly sensitivity construct to estimate future runoff changes (section 4.4) and provide examples for several tributaries (order
1042105 km2). Although the method can be applied to any watershed, we focus on the Willamette
River at Portland (Willamette), Yakima River near Parker (Yakima), Columbia River at Keenleyside Dam
(Upper Columbia), the Snake River at Ice Harbor Dam (Snake), and the Columbia River at The Dalles
(Dalles), subbasins that capture a diversity of responses and are important from a water management
standpoint.
Throughout our comparisons, we consider the nature of modeled hydrologic sensitivities with an
emphasis on two key properties, linearity and superposition. Linearity is used to evaluate the extent to
which sensitivities based on a small change (e.g., 0.1 C) can provide a reasonable approximation of the
sensitivity to a much larger change (e.g., 3 C). Note that the sensitivities are defined as a change in runoff per degree change in forcing. Divergence of the sensitivities to small and large T increases indicates
nonlinearities in the hydrologic system. Superposition is the extent to which hydrologic sensitivities
identified through independent perturbations of P and T can be added to approximate the effects of
changes applied concurrently. Superposition properties determine whether changes applied in different
seasons in independent simulations (e.g., a 0.1 C increase in October to March and a 0.1 C increase in
April to September) can be added together to estimate the joint response (e.g., a 0.1 C annual increase).
When both properties of linearity and superposition hold, the transfer functions we develop are most
applicable.
VANO ET AL.
C 2015. American Geophysical Union. All Rights Reserved.
V
1961
Water Resources Research
10.1002/2014WR015909
2. Site Description
We define the Pacific Northwest (PNW) as the Columbia River basin and its adjacent coastal drainages
(USGS Hydrologic Region 17 plus the Canadian portion of the Columbia). It is a diverse hydroclimatic region
with varied vegetation, soil, and topography [Stanford et al., 2005], with elevations ranging from sea level to
4400 m (Figure 1). Although hydrologic observations are somewhat sparse in the headwaters, climatic and
streamflow data are generally sufficient to evaluate model predictions across most of the hydroclimatic conditions experienced in the region [Elsner and Hamlet, 2010].
The region’s water resources are managed by municipal, county, state, provincial, and federal agencies,
which make spatial information that spans political boundaries particularly valuable. For example, the
Columbia’s headwaters are in Canada, and the Canadian portion (15% of the basin’s drainage area)
accounts for over a third of the river’s flow at the mouth on average [U.S. Army Corps of Engineers (USACE)
and Bonneville Power Administration (BPA), 2012]. Furthermore, the basin’s two largest reservoirs are in Canada. Joint management of the system for flood control and hydropower is controlled by the Columbia River
Treaty of 1964. As this paper is being written, the 2014/2024 Columbia River Treaty Review process is underway with discussion of how the system should be operated in the future [USACE and BPA, 2012]. Among the
concerns is how and where future runoff patterns might change in response to a changing climate.
3. Methods
3.1. Models and Forcing Data Set
We performed a set of control experiments using the Variable Infiltration Capacity (VIC) macroscale landsurface hydrology model [Liang et al., 1994]. We used VIC version 4.0.7 calibrated and validated as in Hamlet
et al. [2010, chapter 5]. This implementation has a higher spatial resolution (1/16 versus 1/8 ) and generally
improved performance relative to earlier implementations [e.g., Hamlet and Lettenmaier, 1999; Payne et al.,
2004]. The increased spatial resolution improves the ability of the model to represent topographic effects
and resolve smaller watersheds and hence provide information relevant to local water management concerns [Elsner and Hamlet, 2010; Wenger et al., 2010].
We forced the VIC model using the same 1/16 latitude-longitude historical model forcing data set developed in previous efforts [Elsner et al., 2010; Hamlet et al., 2010]. The data set includes daily values of P, maximum and minimum air T, and wind speed, which are used in a preprocessing step based on the MTCLIM
algorithms [Thornton and Running, 1999] to calculate other forcing variables such as longwave radiation,
shortwave radiation, and relative humidity (for details see Bohn et al. [2013]). Note that the VIC model as
implemented operates at a 1 h time step when snow is present, and daily otherwise. The MTCLIM algorithms produce hourly model forcings as necessary from the daily P and T max/min inputs. The model was
cold started on 1 January 1915 and spun up for the period 1915–1970. The model state at the end of this
period served as the initial state for all subsequent model experiments. We simulated P and T perturbations
for the period 1970–2004 and calculated sensitivities over the reference period 1975–2004. Our choice of a
relatively recent three-decade period reflects our desire to use a baseline period consistent with recent climate. However, past work has shown that hydrologic model sensitivities do not change much when calculated over different time periods [Vano et al., 2014].
3.2. Hydrologic Sensitivities
We estimated hydrologic sensitivities, precipitation elasticity (e), and temperature sensitivity (S), as
defined in Vano et al. [2012]. We calculated e and S using control experiments where a baseline simulation was compared with a simulation in which either P or T was perturbed. We used the VIC
model because it has been well tested in the region and results from previous studies could be
used in comparisons. The concept is, however, transferable and can be used with other hydrologic
and land-surface models [e.g., Vano et al., 2012]. For the computation of both e and S, changes
were applied for different time periods: year round (every day of the year), in the warm season
(every day in April to September), in the cool season (every day in October to March), and in four 3
month seasons (every day in October to December, January to March, April to June, and July to September referred to as OND, JFM, AMJ, and JAS, respectively). Results were averaged over the 30 year
reference period. We estimated e and S using 1% P increases and 0.1 C increases. We also applied
VANO ET AL.
C 2015. American Geophysical Union. All Rights Reserved.
V
1962
Water Resources Research
10.1002/2014WR015909
3 C increases when determining watershed classifications, which corresponds to the T change projected by 16 GCMs in the latter part of the 21st century [Das et al., 2011], for annual, warm season,
and cool season perturbations. This 3 C increase allowed direct comparison with Das et al. [2011].
We examined the character of hydrologic sensitivities at three spatial scales (grid cell, watershed, and
major subbasin). At the grid level (1/16 or approximately 30 km2), we calculated sensitivities for
each of the 24,108 grid cells in our PNW domain. At the intermediate watershed scale, hydrologic
quantities were aggregated over 226 watershed units (USGS HUC-8 cataloging units in the U.S. and
GeoBase National Hydro Network Work Unit in Canada). The watershed units have an average drainage area of 3000 km2 (ranging from 200 to 11,600 km2). We aggregated runoff from all 1/16 grid
cells that lie at least 50% within these watershed boundaries. Finally, we calculated the aggregate
sensitivities for the Willamette River, Yakima River, Snake River, Upper Columbia River, and Columbia
River at The Dalles as noted in section 1. When calculating hydrologic sensitivities at the grid level
or watershed scale, we simply aggregated the runoff values (sum of the model’s surface runoff and
drainage) over the area of interest. For the major subbasins, we used runoff values routed to a particular stream gage (streamflow) using the routing model of Lohmann et al. [1998] to account for
channel travel times.
We developed two watershed classifications to emphasize differences in S across the region. The first
classification is based on differences in the annual response (total magnitude) to warming applied in
different seasons and includes three categories: (1) locations where annual S is more negative when
warming is applied in the cool season rather than the warm season, (2) locations were the opposite
is true, i.e., annual S is more negative when warming is applied in the warm season rather than the
cool season, and (3) locations where warming applied in the cool season results in positive annual S
values, which is a special case of category two. The second classification identifies locations that are
sensitive to shifts in seasonality. Categories within this classification are based on the difference in S
between the warm season and the cool season in response to cool season warming. Further details
on these two classifications are provided in sections 4.2.3 and 4.2.4, respectively.
3.3. Estimating Future Streamflow
We used the following procedure to construct monthly future hydrographs from monthly e and S
values (employed as transfer functions) in response to seasonal P and T changes. Four sets of
monthly e and S were calculated by applying 1% P and 0.1 C T increases in OND, JFM, AMJ, and
JAS (eight total VIC simulations) where values were relative to the historical simulation in each
month. Regional-average, seasonal GCM changes (DP (%) and DT ( C)) in OND, JFM, AMJ, and JAS
were also calculated between the historical period (1970–1999) and three future periods, 2020s,
2040s, and 2080s (defined as 2010–2029, 2030–2059, 2070–2099, respectively). Future monthly flows
were then constructed based on equation (1)
Qifut 5Qihis
4 Y
11eij DPj
11Sij DTj
(1)
j51
where i is the month (1–12, with month 1 corresponding to October), j is the season (1–4, with season 1 corresponding to OND), Qifut is the future flow in month i, Qifut is the historic flow in month i, DPj and DTj are
changes in season j, and eij and Sij are the precipitation elasticity and temperature sensitivity, respectively, in
month i in response to a change in season j. This equation captures the interaction terms as cross products.
Further details are included in section 4.4.
We compare our seasonal sensitivity-based monthly hydrographs with hydrographs generated using the
more involved full-simulation approach produced for 10 GCMs by Hamlet et al. [2010] (available online at
http://warm.atmos.washington.edu/2860/products/sites) for the five major subbasins introduced in section
1. Hamlet et al. [2010] used the same hydrologic model, historical forcing data, and streamflow routing
scheme as in this study. To generate ‘‘full-simulation’’ values, however, they also downscaled the GCMs
using the hybrid delta method to get daily P and T output to run VIC, after which they routed the streamflow and bias corrected it (for all locations except the Willamette, where only streamflows without bias correction were available).
VANO ET AL.
C 2015. American Geophysical Union. All Rights Reserved.
V
1963
Water Resources Research
10.1002/2014WR015909
4. Results and Discussion
4.1. Spatial Variations in Hydrological Responses
4.1.1. Precipitation Elasticities (e)
To determine the extent to which warm and cool season changes contribute to basin-wide sensitivities, we
applied a 1% P increase year round (Figure 2, left column, top plot), in the warm season (middle), and in the
cool season (bottom). As an example, an e of 3 in the upper left plot of Figure 2 implies that a 1% increase
in P applied throughout the year would result in a 3% increase in annual runoff. For 6 month changes,
superposition holds: the lower two plots, which reflect the contribution of the warm and cool seasons to
annual sensitivities, sum to the top plot to within 60.15% for all grid cells (i.e., to a first approximation,
warm and cool season e values are additive). Seasonally applied P changes reflect the time of year when P
is greatest. In other words, year-round e is primarily determined by cool season changes (on average, more
than 70% of the annual e can be attributed to cool season changes), the season when the majority of P
occurs in the PNW.
Precipitation elasticities (e) exhibit seasonal responses that vary by location due to temperature (mostly
related to snow processes). These seasonal responses are straightforward (and therefore not shown). When
P change is applied in the warm season, essentially all runoff change occurs in the warm season. When P
change is applied in the cool season, runoff changes occur in the cool season at low elevations and in the
warm season at high elevations. However, overall, relative P increases in the
cool season generate more runoff
because cool season P is both more
abundant and more efficient in producing runoff through a large spring pulse
than warm season P.
4.1.2. Temperature Sensitivities (S)
The right column of Figure 2 shows the
sensitivity of annual responses to warming applied annually, in the warm and in
the cool season. Superposition applies in
most grid cells. That is, the sum of the
responses to warm and cool season
warming is close to the response to
year-round warming (within 60.25% per
C for 94% of grid cells). Spatially, there
is a range of responses including both
increases (blue) and decreases (red) in
annual runoff. If warming occurs only in
the cool season, 74% of the grid cells
show decreases in annual runoff,
whereas if warming occurs throughout
the year 93% of the grid cells show
decreases in annual runoff. The larger
number of cells with decreases results
because declines from warm season
warming are greater than increases from
cool season warming for most grid cells.
Figure 2. Annual responses to (left) P (e values) and (right) T (S values) change.
Breaks in color bars indicate changes in increment.
VANO ET AL.
C 2015. American Geophysical Union. All Rights Reserved.
V
The mechanism by which runoff
increases in response to warming is
sometimes referred to as the Dettinger
hypothesis, described in Jeton et al.
[1996]: due to warming, water leaves
the system earlier in the year (as a result
of earlier snowmelt or because P falls as
1964
Water Resources Research
10.1002/2014WR015909
Figure 3. Temperature linearity, effects of T increment on annual temperature sensitivities (S). S values (ranging from 216% to 115%)
from simulations using a (left) 0.1 C increase and a (right) 3 C increase and differences between the two. Differences in S less than one are
shown in white (right plot).
rain rather than snow) and is not available for evapotranspiration later in the year. Therefore, it is possible
for runoff to be higher in a warmer climate, even without a change in P.
In aggregate, we found the annual response was greater from warming applied in the warm season than
warming applied in the cool season, which is similar to Das et al. [2011] values at The Dalles, but local
watershed-level responses varied (discussed in section 4.2.2). Note that Das et al. [2011] evaluated the basin
response to a 3 C increase. This response differs from an empirical analysis by Berghuijs et al. [2014] of 97
small U.S. catchments that showed greater streamflow declines from temperature changes in November to
March than in May to September, highlighting an area where more mechanistic work would be helpful.
To test the effect of the different change increment, and hence the linearity of annual S values, we compared the patterns of annual S values across the PNW using a 0.1 C increment with those generated using a
3 C increment (Figure 3). The patterns are similar and differences between a 0.1 C and 3 C perturbation are
less than 1% in 78% of the grid cells (and less than 2% in 95% of grid cells). Sensitivities calculated with the
smaller perturbation (0.1 C) are generally greater in absolute value (both larger increases and decreases in
runoff) than those calculated with the 3 C perturbation. While larger T changes have the potential to melt
more snow, this snow may not be available. In the case that all the snowmelts for a T increase of less than
3 C, the sensitivities in response to a 3 C perturbation are smaller than those for a 0.1 C (note that all sensitivities are expressed per C warming).
Figure 4 shows how warming imposed annually and in each season (rows) contributes to changes in individual
seasons, where the left column is the annual response (same three plots as in Figure 2), the middle column is
the change in the warm season, and right column is the change in the cool season. Values are calculated relative to annual runoff. Therefore, by definition, the warm and cool season S add to the annual S (horizontally,
denominators are the same). Warm season warming reduces runoff throughout the year, whereas cool season
warming increases runoff in the cool season but reduces runoff in the warm season. Therefore, although the
net runoff change is less for warming in the cool season than in the warm season, the change in seasonality is
much greater for cool season warming. The differences in hydrologic responses in warm and cool seasons are
largely controlled by regions at higher elevations (identified in Figure 1). This is particularly important as not all
seasons show equal warming [Abatzoglou et al., 2014]. Future projections of seasonal change in the PNW indicate greater warming in the summer than other seasons [Mote et al., 2013].
4.2. Categorizing Hydrologic Changes at the Watershed Scale
4.2.1. Annual Watershed-Level Precipitation Elasticities (e) and Temperature Sensitivities (S)
We complement the fine-scale spatial information in Figures 2 and 4 with watershed-level categorizations.
Figure 5 shows the watershed-scale response to annual P and T changes. This information is similar to the
top plots in Figure 2, but is averaged for each of the 226 watersheds in the region. Watershed annual e
VANO ET AL.
C 2015. American Geophysical Union. All Rights Reserved.
V
1965
Water Resources Research
10.1002/2014WR015909
Figure 4. Seasonal responses to warming. (left) Annual sensitivities (% change in annual runoff per C warming) and the contribution of
that change in the (center) warm season and (right) cool season for warming applied throughout (top) the year; (middle) only in the warm
season; and (bottom) only in the cool season. The annual sensitivities (left plots) contain the same information as annual sensitivity values
in Figure 2. The breaks in the color bar indicate a change in increments, emphasizing the larger absolute values of warm and cool season
sensitivities.
values range from 0.9 to 2.5. Five watersheds have an e of less than one. These locations have low annual P
and are located at lower elevations in eastern Washington and the Middle Snake subbasin. Watersheds with
higher e values are located in west-central Idaho in midelevation to high-elevation regions where model
simulations show relatively low runoff ratios. These annual values can be compared with results from Sankarasubramanian et al. [2001] who evaluated precipitation elasticities from annual P and streamflow observations throughout the conterminous U.S. using a nonparametric observation-based estimator of elasticity.
Within the PNW, the map provided in Sankarasubramanian et al. [2001] has contours with magnitudes ranging from 1.0 to 2.0, with a similar increase in elasticities from west to east, although they report higher values in eastern Oregon, rather than west-central Idaho. Their continental-scale analysis is at a coarser spatial
resolution than ours (dictated by station locations), which may explain why specific locations do not align.
Regions with the greatest annual S values in response to annually applied T changes (Figure 5, right) are
also clustered in the upper Salmon River basin of central Idaho, while less sensitive areas are at both
extremes of the elevation spectrum—low lying coastal areas (rain-dominated watersheds) and headwaters
in Canada and Wyoming (snow-dominated watersheds). All watersheds show decreases in runoff with
increasing T (S values range from 20.2% to 28% per C), i.e., increases in runoff represented by the blue
grids in Figure 4 top left, are outweighed by decreases when averaged with the other grids in the
VANO ET AL.
C 2015. American Geophysical Union. All Rights Reserved.
V
1966
Water Resources Research
10.1002/2014WR015909
Figure 5. Watershed-scale annual precipitation elasticities (e) and temperature sensitivities (S). Annual responses (percent) to annually
applied change (percent precipitation change or C warming).
watershed. These annual responses to annually applied changes capture one aspect of S. In the next section,
we discuss characterizations that address seasonality.
4.2.2. Annual Responses to Warm and Cool Season Warming (Total Magnitude)
Figure 6a classifies the 226 watersheds into one of three categories according to the annual sensitivity to
warm and cool season warming (using a 3 C change increment):
1. More Sensitive to Cool Season Warming. Locations where annual runoff declines more with cool season
warming than with warm season warming. 83 of the 226 subbasins (36%), mostly low elevation watersheds, have this characteristic, which is similar to temperature sensitivities in the southern Sierra shown
in Das et al. [2011].
2. More Sensitive to Warm Season Warming. Annual runoff magnitudes decline more with warm season
warming than with cool season warming. This characteristic is seen in basin-wide responses in the Colorado, Columbia, and North Sierra [Das et al., 2011]. This response is most common (in 64% of watersheds,
including the special case below), occurring for watersheds at midelevation and high elevation and at
higher latitudes along the coast.
3. Cool Season Warming Positive. In this special case, warming applied in the cool season generates
increases in cool season runoff that are greater than decreases in warm season runoff resulting in
net increases in runoff (i.e., warmer winter temperatures lead to increased annual flows). This occurs
in 14% of watersheds in the PNW (Figure 6a), most of which are at high elevations (see Figure 1 for
elevations).
In general, even though the Columbia River basin as a whole has substantially greater sensitivity of annual
runoff to warm rather than cool season warming (category 2), 27% of the 151 watersheds upstream of The
Dalles show the reverse behavior, being more sensitive to cool season warming (category 1). These basins
are mostly at lower elevations in Oregon, eastern Washington, and southern Idaho. This pattern pertains to
many of the coastal drainages downstream of The Dalles as well, especially in Oregon (Figure 6a, yellow
watersheds).
4.2.3. Seasonal Responses to Warming (Shifts in Seasonality)
To quantify the sensitivity of watersheds to shifts in seasonality, we defined six categories that are gradations of the difference between cool season and warm season sensitivities to cool season warming
VANO ET AL.
C 2015. American Geophysical Union. All Rights Reserved.
V
1967
Water Resources Research
10.1002/2014WR015909
Figure 6. PNW watershed classifications. (a) Watershed classifications based on temperature sensitivities to warm and cool season warming of 3 C that reflect the time of year having the greatest impact on annual flow magnitudes; (b) classifications based on difference in
percentage points between warm and cool season S when warming is applied in the cool season. Larger differences (dark orange) are
more likely to experience seasonal differences in their hydrographs as temperatures increase.
for all PNW watersheds (Figure 6b). These differences are similar to differences in percentage points
between the middle and right columns of Figure 4 for change applied in the cool season. Watersheds
having the largest differences (dark orange) are watersheds most sensitive to warming since, in this
region, the difference is largely due to runoff increasing in the cool season and subsequently decreasing in the warm season. These dark orange watersheds correspond to watersheds that change from
snow-dominated to mixed snow/rain systems in future simulations as seen in Elsner et al. [2010] and
Tohver et al., [2014]. In other words, watersheds that change from one regime (e.g., snow-dominated)
to another (e.g., transient) are also those that exhibit the greatest seasonal shifts in runoff (largest differences) in response to cool season T perturbations. Both snow-dominated, high-elevation watersheds
(hydrographs peak in the spring), as well as rain-dominated, low-elevation watersheds (hydrographs
peak in the fall-winter) do not show large differences in their warm and cool season S. For these
basins, warming does not result in a regime change because they remain cold enough that the
applied warming does not affect the timing of their snowmelt, or because they are already rain dominated and so there is little snow to melt.
4.3. Monthly Responses to Seasonal Precipitation and Temperature Changes
To quantify monthly responses, we applied changes in 3-month periods (OND, JFM, AMJ, and JAS), calculated monthly responses (supporting information Tables A1 and A2), and created ‘‘bubble’’ diagrams (Figure
7) similar to those in Nijssen et al. [2001]. These bubble diagrams show how changes applied in one season
(y axis) effect changes in streamflow in every month of the year (x axis, starting in October) for five major
subbasins.
Figure 7 shows eij and Sij , where j is the season in which the change is applied (y axis) and i is the month in
which the resulting streamflow change is evaluated (x axis; also see equation (1)). For example, e12
2 is the
elasticity of September streamflow in response to a 1% change in P imposed during the JFM season. Values
across the top row (marked ‘‘ALL’’) are calculated from simulations where the change is applied in every
month of the year whereas the individual rows are based on simulations in which changes are imposed in
an individual season. In general, superposition holds, meaning changes are additive, i.e., the sum of the four
3-month rows equals the ‘‘ALL’’ row, but for reasons discussed below, this is not always the case. The rightmost column (marked ‘‘Y’’) shows the annual change, which is the total difference throughout the year
divided by the annual streamflow (note: values do not add horizontally as they are percentages for individual months, i.e., denominators differ).
Precipitation elasticity values (Figure 7, middle) for four of the basins (all except the Willamette) are similar
because P changes applied in OND, JFM, and AMJ all affect streamflow response throughout the summer. This
VANO ET AL.
C 2015. American Geophysical Union. All Rights Reserved.
V
1968
Water Resources Research
10.1002/2014WR015909
Figure 7. Monthly responses to seasonal P (e, % change streamflow per % change P) and T changes (S, % change streamflow per C). Warming is applied in 3-month periods (OND, JFM,
AMJ, JAS) and throughout the year (ALL), and responses (e or S) are shown for each individual month relative to that month (O, N, D, . . . S) as well as the annual response (Y). Maps on
the left show the contributing area for each tributary.
reflects the nature of large tributaries in the Columbia River basin. All are snowmelt dominated (see historical
hydrographs in Figure 10), illustrating the dependence of the seasonal P response on seasonal T (as discussed
in section 4.1.1). Only rain-dominated basins, typically located on the west side of the Cascades, have monthly
e values where the majority of the streamflow changes occur in the same months as the P perturbation. The
Willamette, which has a greater peak from rain in the fall and winter than from snowmelt in the spring, is an
example of a basin where responses are smaller (because of less snow) and correspond more directly with the
timing of P changes.
The bubble diagrams for S (Figure 7, right) show how streamflow responses to warming vary by basin. The
larger blue (increases) and red (decreases) bubbles highlight basins that are most sensitive to change (e.g.,
Yakima) and the time of year when S transitions from positive to negative. For example, the largest bubbles
VANO ET AL.
C 2015. American Geophysical Union. All Rights Reserved.
V
1969
Water Resources Research
10.1002/2014WR015909
in the Willamette (red bubbles in JJA
on the x axis) indicate decreases in
summer streamflow in response to
warming in all seasons, while winter
streamflow is relatively insensitive to
warming (blue bubbles from November to March are small relative to
other basins). Note that the bubble
diagrams show changes relative to
each month so the absolute changes
in winter may still be larger than in
summer. The Yakima, Upper Columbia,
and Snake are all currently snow
dominated, but the Yakima is more
vulnerable to T increases in JFM than
the other basins and, based on this
analysis, could experience a greater
change in seasonality as T increases.
The Upper Columbia and Snake are
more affected by AMJ warming (see
AMJ row relative to others) as higher
Figure 8. Monthly responses for (top) precipitation elasticities, e, and (bottom)
elevations (hence lower T) provide a
temperature sensitivities, S, in the Yakima River basin calculated using different
buffer such that these basins are relachange increments (colored lines). The largest differences (greatest nonlinearities)
occur in the summer time, when lines diverge.
tively insensitive to winter warming.
The Dalles reflects the combined basin
response, but the domain is essentially overweighted by the areas contributing the most runoff; hence, this
domain is most similar to the high-elevation headwater tributaries.
4.3.1. Linearity of Monthly Responses to Seasonal Change
We test the linearity of monthly responses by varying the increment of annually applied change used to
calculate e and S values in the Yakima (Figure 8). If the response were linear, e and S values would be
the same (note that e and S are normalized to represent the change in streamflow for 11% P or 11 C T,
respectively). For e values, we applied increments of 220%, 210%, 10.1%, 110%, and 120%. For this
range of increments, winter and spring responses are approximately linear, while summer and fall
responses are not (the largest range in seasonal e values is in August, which coincides with low flows).
For S values, we applied increments of 10.1, 13, and 16 C. Again, streamflow responds similarly during
certain months regardless of the increment (e.g., October and November), but considerable differences
exist, especially in the summer. The large differences in the summer correspond to the time of year
when streamflow is historically low, and relatively small changes in magnitude can have a large effect on
e and S.
4.3.2. Superposition of Monthly Responses to Seasonal Change
For P changes, superposition generally applies, that is, monthly e values are close to additive (supporting
information Table A1). Values of e for each of the four 3-month periods (the OND, JFM, AMJ, JAS rows) calculated with a 1% P change in those seasons, closely sum to e values calculated from a 1% P change applied
annually (the ALL row). The absolute difference is less than 1% for all but 1 month at one location. The Yakima in September shows a larger difference (24.7%), which corresponds to the lowest flow month when
small absolute changes result in larger relative changes.
For T changes, superposition holds less uniformly. Annual S values for seasonally applied T changes are generally additive, whereas monthly S values for seasonally applied T changes, especially for large T increments,
are not (supporting information Tables A2 and A3). For example, in the Yakima, the annual S value calculated based on a 0.1 C increase in temperature applied year round is nearly identical to the annual S value
calculated as the sum of the responses to a 0.1 C increase applied in each of the four seasons separately
(22.667% versus 22.672% per C, respectively) (supporting information Table A2). Similarly, annual S values
VANO ET AL.
C 2015. American Geophysical Union. All Rights Reserved.
V
1970
Water Resources Research
10.1002/2014WR015909
calculated in these two ways based on a 3 C increase differ by only 0.37% per C (supporting information
Table A3).
In contrast, whether monthly S values adhere to principles of superposition depends on the increment of
change. In the Yakima, monthly S values calculated with a 10.1 C change are still mostly additive: the sum
of the four seasons is only 1.7% per C lower in August and 1.6% per C higher in September than monthly
S values calculated from annually applied T change (supporting information Table A2). Alternatively,
monthly S values calculated with an increment of 13 C range from the sum of the four seasons being
18.2% points lower in July (243.5% versus 225.3%) to 13.5% points higher in April (15.6% versus 27.9%)
(supporting information Table A3). Differences in monthly S values become even greater if changes are
applied in individual months instead of 3-month intervals (not shown). This nonadditive response can be
attributed to the influence of snowpack and the fact that T increases, if applied in individual seasons and
then added together, can effectively ‘‘melt the same snow twice,’’ resulting in estimates of higher springtime and lower summertime streamflow. This highlights the importance of selecting change increments
that approximate the tangent of the change function when superposition will be applied (as in section 4.4).
Of the five locations simulated, the Yakima has the largest change in seasonal streamflow with T increases,
and thus is most sensitive to these T effects.
4.4. Application to Climate Change Projections
A practical application of the seasonal hydrologic sensitivities outlined above is to use monthly e and S values in response to seasonal change as transfer functions to convert seasonal changes in P and T from GCMs
into future projections of monthly hydrographs. Although the method is approximate and only provides an
estimate of changes in the mean monthly streamflow (and not, for instance, changes in variability or a multiyear time series), it has the advantage that it is much less time consuming than full-simulation methods. In
this section, we illustrate how the seasonal sensitivity-based method compares with the full-simulation
method of Hamlet et al. [2010].
Seasonal sensitivity-based hydrographs were calculated by applying P and T changes from GCM simulations
for each 3-month season (equation (1), see supporting information and equation (A1) for an alternative
form of the equation). For the Yakima, we evaluated how our results were affected by the inclusion (equation (1)) or exclusion (equation (A2) in supporting information) of the interaction terms (Table 1). The interaction terms represent the compounded effects of changes in P and T in one season on the runoff in other
seasons, for example S1OND DTOND S1JFM DTJFM . The Yakima basin has the greatest seasonal sensitivity to warming of the five basins and would be most sensitive to this change. We found that including seasonal interaction terms is most important for T changes in this basin. If P interactions are not included, results change
only slightly. If future seasonal P and T changes were applied without interaction terms (supporting information equation (A2)), annual responses are closest to full-simulation results, but the seasonal sensitivitybased hydrographs significantly overestimate reductions in summertime streamflow compared to fullsimulation hydrographs. For example, streamflow becomes negative in the Yakima in July for some scenarios in 2040 and all scenarios in 2080 (Table 1). The inclusion of the interaction terms reduces summertime
declines by applying the change fractions to values that have been reduced according to T impacts in other
seasons. In the same way, wintertime streamflow can be inflated. Nonetheless, negative streamflow values
can result as changes are being applied in a linear manner when, in reality, sensitivities are nonlinear. For
example, if there is a 4 C increase in AMJ when July’s sensitivity is 229% per C, the percent decline will be
over 100%. Consequently, this estimation method works best for small changes (e.g., more typically 30
years into the future rather than 100) and for basins with modest sensitivities, where nonlinearities in the
sensitivities are small. Overall, this method is useful only for the seasonal nature of hydrograph changes
rather than the absolute magnitude of future streamflow. The latter, either annually or in a particular season, should be viewed with caution.
Figure 9 and Table 1 show results for the Yakima for three future time periods. Generally, the seasonal
sensitivity-based method reproduces the change in seasonality quite well, providing similar information
about both the average seasonal response (average of the 10 GCMs) and ensemble range, even as climate
change becomes more pronounced in 2080. This method best captures the near-term changes, which are
also of greater importance to water managers for 30 year planning horizons. As expected, the ensembles
(gray lines) of the seasonal sensitivity-based method are smoother than those from the full-simulation
VANO ET AL.
C 2015. American Geophysical Union. All Rights Reserved.
V
1971
Water Resources Research
10.1002/2014WR015909
Table 1. Yakima Scenarios, Differences From Historical Streamflow
% Difference in Annual Magnitudes
Full-simulation [Hamlet et al., 2010]
P and T interactions (Figure. 8 and 9)
No interactions
T interactions only
% Increases in Winter (November to April)
Full-simulation [Hamlet et al., 2010]
P and T interactions (Figures 8 and 9)
No interactions
T interactions only
% Decrease in Peak Flow (May)
Full-simulation [Hamlet et al., 2010]
P and T interactions (Figures 8 and 9)
No interactions
T interactions only
% Decrease in Low Flow (July)
Full-simulation [Hamlet et al., 2010]
P and T interactions (Figures 8 and 9)
No interactions
T interactions only
2020
2040
2080
2% (211 to 15%)
20% (214 to 18%)
21% (216 to 16%)
20% (215 to 17%)
5% (215 to 22%)
5% (26 to 24%)
2% (210 to 19%)
5% (26 to 23%)
6% (212 to 31%)
14% (212 to 47%)
5% (221 to 30%)
14% (213 to 45%)
30% (5 to 53%)
22% (2 to 55%)
22% (4 to 50%)
22% (2 to 55%)
48% (28 to 87%)
44% (27 to 82%)
43% (29 to 71%)
44% (27 to 81%)
72% (39 to 127%)
83% (39 to 153%)
79% (49 to 122%)
83% (38 to 149%)
213% (221 to 2%)
29% (220 to 1%)
28% (221 to 1%)
210% (220 to 1%)
225% (255 to 1%)
216% (226 to 4%)
212% (222 to 6%)
216% (226 to 4%)
251% (279 to 21%)
235% (264 to 0%)
226% (253 to 7%)
235% (264 to 1%)
245% (263 to 230%)
249% (269 to 230%)
257% (289 to 230%)
250% (269 to 230%)
263% (286 to 243%)
271% (294 to 241%)
289% (2142 to 237%)
271% (294 to 241%)
284% (294 to 268%)
297% (2112 to 277%)
2157% (2257 to 286%)
2157% (2257 to 286%)
a
Transfer functions were applied to seasonal values of P and T changes for each GCM. Differences are between historical streamflows
(1970–1999) and three future periods (2010–2029, 2030–2059, 2070–2099). Values are the average and range of 10 GCMs for the A1B
emissions scenario, the same ones simulated by Hamlet et al. [2010], also shown in this table.
Figure 9. Yakima River basin at Parker monthly average streamflow projections for the 2020s, 2040s, and 2080s for the A1B emissions scenario. The left plots are results from Hamlet et al. [2010], which use a full-simulation method, whereas the middle plots are generated using
the seasonal sensitivity-based method presented in this study. The right plots compare averages of the 10 GCMs for both methods.
VANO ET AL.
C 2015. American Geophysical Union. All Rights Reserved.
V
1972
Water Resources Research
10.1002/2014WR015909
Figure 10. Comparison of monthly average streamflow projections similar to Figure 9 but for only the 2040s at five locations throughout the Columbia River basin.
method, since they are based on changes applied to long-term averages. For summertime streamflow, the
seasonal sensitivity-based method captures the nature of the changes, but actual magnitudes are not realistic for reasons mentioned earlier.
We compare results between these two methods in five locations (Figure 10), across the spectrum of
rain-dominated (Willamette) to snow-dominated basins (Upper Columbia), as indicated by the season
of peak streamflow. We present results from 2040 as it tests the method more than 2020, but is not
as extreme as 2080, and provides more information for near-term water management planning. Results
show that across the spectrum, the seasonal sensitivity-based method captures the nature of seasonal
changes, both in the average and range of seasonal responses. The greatest difference between the
two methods is the rate at which streamflow declines between June and September. This relates to
VANO ET AL.
C 2015. American Geophysical Union. All Rights Reserved.
V
1973
Water Resources Research
10.1002/2014WR015909
the nonlinearity in the seasonal response, which is greater in the summer (as highlighted in Figure 8).
Additionally, the seasonal sensitivity-based approach is most vulnerable to error when S values are
large and negative, which also occurs in the summer (especially in July). In all cases, and especially in
locations with the coldest temperatures (e.g., Upper Columbia), the seasonal sensitivity-based streamflow declines faster than that of the full-simulation streamflow. In the bubble diagram for the Upper
Columbia, this is seen as the contrast between the positive S in June and the large negative S in July;
when change is continuous, as in the full-simulation, this transition is less abrupt. Overall, summertime
streamflow with the seasonal sensitivity-based method is lower (a more conservative estimate) than
with the full-simulation method.
In general, these results support a straightforward approach to estimating changes in monthly streamflow
based on future changes in monthly P and T from GCMs. The approach uses only two sets of 48 values each
(12 monthly values times 4 seasons), one for e and one for S, as transfer functions to reproduce seasonal
changes created from detailed model simulations. This simplified approach provides plausible monthly
hydrographs, but has important limitations. Its utility and accuracy are greater for applications of seasonal
shifts, not changes in total magnitude. This study has focused on seasonal shifts because water management in the PNW is more concerned with climate-forced seasonalities (due to small reservoir capacity relative to annual streamflows) than with total magnitude changes. If, however, magnitudes are more
important, the alternate sensitivity-based technique that emphasizes adjustments designed to capture
annual responses [Vano and Lettenmaier, 2014] may be more appropriate. Another limitation is that the set
of e and S values presented here are specific to the baseline T and current land cover in the historical hydrologic model simulations. If these change then the sensitivities will likely also change. Changing T and land
cover are less of a problem in the near term and are also not reflected in most full-simulation approaches.
Updating e and S values would require nine model runs, which is a considerably less demanding computational effort than that required for full-simulation estimations (number of GCMs 3 number of emission scenarios 3 number of future time periods). Generally, the seasonal sensitivity-based method described and
tested here uses linear approximation of seasonal e and S, which in fact are nonlinear. This typically results
in overestimates of streamflow changes, especially in summer, and an exaggeration of the transition from
the cool to the warm season.
5. Conclusions
The Pacific Northwest (PNW) will show different responses to climate change depending on the specific
location, season, and magnitude of P and T change. To better define how the region, and its local watersheds, will respond to future change, we map the sensitivities of runoff to modeled changes in P and T
both spatially and temporally. We use concepts of hydrologic sensitivities to (a) quantify watershed classifications that help define local hydrologic changes and (b) develop a methodology where sensitivities serve
as a transfer function that can be used to quickly convert average seasonal P and T changes into an estimate of future hydrographs for local watersheds.
Through application of the hydrologic sensitivity approach to the PNW we found:
1. Transitional (intermediate elevation) watersheds experience the greatest seasonal shifts in runoff in
response to cool season warming.
2. For watersheds that receive much of their winter P as snow (i.e., snow-dominated and transitional watersheds), warming applied in the cool season, results in both greater change in the seasonality of runoff
and smaller reductions in annual runoff compared to warming applied in the warm season.
3. The net change from cool season warming results in annual runoff increases for 26% and decreases for
74% of the domain. Runoff increases occur in the high-elevation regions in primarily snow-dominated
watersheds but are overwhelmed by decreases when averaged over larger areas or when warming is also
applied in the warm season.
When applying hydrologic sensitivities to future change, we found:
1. The ability to construct hydrographs from monthly hydrologic sensitivities depends on how well hydrologic sensitivities adhere to principles of linearity and superposition. For this reason, the seasonal
sensitivity-based method works best for modest changes (e.g., 30 years into the future rather than 100,
VANO ET AL.
C 2015. American Geophysical Union. All Rights Reserved.
V
1974
Water Resources Research
10.1002/2014WR015909
which is also a more typical planning horizon for management agencies) and for basins where sensitivities are modest, hence nonlinearities in the sensitivities are small.
2. Seasonal P changes (and their monthly e values calculated using an increment of 1%) are largely independent of each other (e.g., the runoff response from P increases in JFM and AMJ generally have minimal
interaction), adhering to principles of superposition.
3. Seasonal T changes (and their monthly S values calculated using a 10.1 C increment) also appear to be
additive, although less so than for P. We found 3-month seasonal T changes based on a 10.1 C increment resulted in largely independent responses. However, if the increment of change is larger (i.e.,
13 C), superposition does not hold. In the case of 13 C, the simple approach can essentially double
count the response by melting the same snow twice.
4. Future changes in T should be incorporated using an approach that accounts for interactions between
seasons. This results in estimated future hydrographs that more closely replicate hydrographs generated
using the full-simulation approach. Conversely, interactions between seasons for P changes have little
effect on the estimated future hydrograph.
5. The seasonal sensitivity-based method compares well with the full-simulation method in these basins for
purposes of capturing the nature of changes in seasonality. There are, however, seasons and locations
where P and T changes are more nonlinear; and caution should be used in interpreting the absolute
streamflow magnitude (rather than the relative change), especially in the summer and when future T
increases are large.
Acknowledgments
The authors thank Alan Hamlet, David
Pierce, Tapash Das, and Daniel Cayan
for their insights that led to this work
and Marketa Elsner and Se-Yeun Lee for
their assistance in model setup. The
manuscript was improved from
thoughtful comments by the Associate
Editor, Patrick Kormos, and two
anonymous reviewers. Forcing data
and future simulation streamflow
values were provided by the Columbia
Basin Climate Change Scenarios Project,
University of Washington, http://www.
hydro.washington.edu/2860, and GCM
precipitation and temperature changes
where calculated from data
downloaded at http://esg.llnl.gov:8080/
home/publicHomePage.do. Support for
this work was provided by NOAA’s
Regional Integrated Science
Assessment program, grant
NA10OAR4310218, to Oregon State
University’s Climate Impacts Research
Consortium and the National Science
Foundation under Award EAR-1250087
to the first author.
VANO ET AL.
References
Abatzoglou, J. T., and T. J. Brown (2012), A comparison of statistical downscaling methods suited for wildfire applications, Int. J. Climatol.,
32, 772–780, doi:10.1002/joc.2312.
Abatzoglou, J. T., D. E. Rupp, and P. W. Mote (2014), Understanding seasonal climate variability and change in the Pacific Northwest of the
United States, J. Clim., 27, 2125–2142, doi:10.1175/JCLI-D-13-00218.1.
Berghuijs, W. R., R. A. Woods, and M. Hrachowitz (2014), A precipitation shift from snow towards rain leads to a decrease in streamflow,
Nat. Clim. Change, 4(7), 583–586.
Bohn, T. J., B. Livneh, J. W. Oyler, S. W. Running, B. Nijssen, and D. P. Lettenmaier (2013), Global evaluation of MTCLIM and related algorithms for forcing of ecological and hydrological models, Agric. For. Meteorol., 176, 38–49, doi:10.1016/j.agrformet.2013.03.003.
Das, T., D. W. Pierce, D. R. Cayan, J. A. Vano, and D. P. Lettenmaier (2011), The importance of warm season warming to western U.S. streamflow changes, Geophys. Res. Lett., 38, L23403, doi:10.1029/2011GL049660.
Dooge, J. C., M. Bruen, and B. Parmentier (1999), A simple model for estimating the sensitivity of runoff to long-term changes in precipitation without a change in vegetation, Adv. Water Resour., 23, 153–163, doi:10.1016/S0309-1708(99)00019-6.
Dooge, J. C. I. (1992), Sensitivity of runoff to climate change: A Hortonian approach, Bull. Am. Meteorol. Soc., 73, 2013–2024.
Elsner, M. M., and A. F. Hamlet (2010), Macro-scale hydrologic model implementation, report, University of Washington, Seattle, Wash.
[Available at http://warm.atmos.washington.edu/2860/r7climate/study_report/CBCCSP_chap5_vic_final.pdf.].
Elsner, M. M., L. Cuo, N. Voisin, J. S. Deems, A. F. Hamlet, J. A. Vano, K. E. B. Mickelson, S. Y. Lee, and D. P. Lettenmaier (2010), Implications of
21st century climate change for the hydrology of Washington State Climatic Change, Clim. Change, 102(1), 225–260.
Federal Columbia River Power System (FCRPS) (2001), The Columbia River System Inside Story: Federal Columbia River Power System, 2nd ed.,
Bonneville Power Admin., Portland, Oreg.
Fu, G., S. P. Charles, and F. H. S. Chiew (2007), A two-parameter climate elasticity of streamflow index to assess climate change effects on
annual streamflow, Water Resour. Res., 43, W11419, doi:10.1029/2007WR005890.
Hamlet, A. F. (2011), Assessing water resources adaptive capacity to climate change impacts in the Pacific Northwest Region of North
America, Hydrol. Earth Syst. Sci., 15(5), 1427–1443.
Hamlet, A. F., and D. P. Lettenmaier (1999), Effects of climate change on hydrology and water resources in the Columbia River basin, J. Am.
Water Resour. Assoc., 35(6), 1597–1623.
Hamlet, A. F., et al. (2010), Final project report for the Columbia Basin climate change scenarios project, report, University of Washington,
Seattle, WA. [Available at http://www.hydro.washington.edu/2860/report/.].
Hamlet, A. F., M. M. Elsner, G. S. Mauger, S. Y. Lee, I. Tohver, and R. A. Norheim (2013), An overview of the Columbia basin climate change
scenarios project: Approach, methods, and summary of key results, Atmos. Ocean, 51(4), 392–415, doi:10.1080/07055900.2013.819555.
Jeton, A. E., M. D. Dettinger, and J. L. Smith (1996), Potential effects of climate change on streamflow, eastern and western slopes of the
Sierra Nevada, California and Nevada, U.S. Geol. Surv. Water Resour. Invest. Rep., 95–4260, 44 pp.
Lee, S. Y., A. F. Hamlet, C. J. Fitzgerald, and S. J. Burges (2009), Optimized flood control in the Columbia river basin for a global warming scenario, J. Water Resour. Plann. Manage., 135(6) 440–450, doi:10.1061/(ASCE)0733-9496(2009).
Liang, X., D. P. Lettenmaier, E. F. Wood, and S. J. Burges (1994), A simple hydrologically based model of land surface water and energy
fluxes for general circulation models, J. Geophys. Res., 99, 14,415–14,428.
Lohmann, D., E. Raschke, B. Nijssen, and D. P. Lettenmaier (1998), Regional scale hydrology: I. Formulation of the VIC-2L model coupled to
a routing model, Hydrol. Sci. J., 43(1), 131–141.
Luce, C. H., and Z. A. Holden (2009), Declining annual streamflow distributions in the Pacific Northwest United States, 1948–2006, Geophys.
Res. Lett., 36, L16401, doi:10.1029/2009GL039407.
Luce, C. H., J. T. Abatzoglou, and Z. A. Holden (2013), The missing mountain water: Slower westerlies decrease orographic enhancement in
the Pacific Northwest USA, Science, 342(6164), 1360–1364.
C 2015. American Geophysical Union. All Rights Reserved.
V
1975
Water Resources Research
10.1002/2014WR015909
Mantua, N., I. Tohver, and A. F. Hamlet (2010), Climate change impacts on streamflow extremes and summertime stream temperature and
their possible consequences for freshwater salmon habitat in Washington State, Clim. Change, 102(1–2), 187–223, doi:10.1007/s10584010-9845-2.
Milly, P. C. D. (1992), Potential evaporation and soil moisture in general circulation models, J. Clim., 5(3), 209–226.
Mote, P., J. Abatzoglou, and K. Kunkel (2013), Climate change in the Northwest, in Climate Change in the Northwest: Implications for Our
Landscapes, Waters, and Communities, edited by M. Dalton, P. W. Mote, and A. K. Snover, chap. 2, 224 pp., Island Press, Washington D. C.
Nijssen, B., G. M. O’Donnell, A. F. Hamlet, and D. P. Lettenmaier (2001), Hydrologic sensitivity of global rivers to climate change, Clim.
Change, 50(1), 143–175.
Payne, J. T., A. W. Wood, A. F. Hamlet, R. N. Palmer, and D. P. Lettenmaier (2004), Mitigating the effects of climate change on the water
resources of the Columbia River basin, Clim. Change, 62, 233–256.
Safeeq, M., G. E. Grant, S. L. Lewis, M. G. Kramer, and B. Staab (2014), A hydrogeologic framework for characterizing summer streamflow
sensitivity to climate warming in the Pacific Northwest, USA, Hydrol. Earth Syst. Sci., 18(9), 3693–3710.
Sankarasubramanian, A., R. M. Vogel, and J. F. Limbrunner (2001), Climate elasticity of streamflow in the United States, Water Resour. Res.,
37, 1771–1781.
Schaake, J. C. (1990), From climate to flow, in Climate Change and U.S. Water Resources, edited by P. E. Waggoner, pp. 177–206, John Wiley, N. Y.
Stanford, J. A., F. R. Hauer, S. V. Gregory, and E. B. Snyder (2005), The Columbia River, in Rivers of North America, edited by A. C. Benke and
C. E. Cushing, Elsevier, Boston, Mass.
Thornton, P. E., and S. W. Running (1999), An improved algorithm for estimating incident daily solar radiation from measurements of temperature, humidity, and precipitation, Agric. For. Meteorol., 93, 211–228.
Tohver, I. M., A. F. Hamlet, and S.Y. Lee (2014), Impacts of 21st-Century Climate Change on Hydrologic Extremes in the Pacific Northwest
Region of North America. J. of the American Water Resour. Assoc. 50(6), 1461–1476, doi: 10.1111/jawr.12199.
U.S. Army Corps of Engineers (USACE) and Bonneville Power Administration (BPA) (2012), Columbia River Treaty, Portland, Oreg. [Available
at http://www.crt2014-2024review.gov/, last accessed 15 June 2012.]
Vano, J. A., and D. P. Lettenmaier (2014), A sensitivity-based approach to evaluating future changes in Colorado River Discharge, Clim.
Change, 122(4), 621–634, doi:10.1007/s10584-013-1023-x.
Vano, J. A., M. Scott, N. Voisin, C. O. Stockle, A. F. Hamlet, K. E. B. Mickelson, M. M. Elsner, and D. P. Lettenmaier (2010), Climate change
impacts on water management and irrigated agriculture in the Yakima River basin, Washington, USA, Clim. Change, 102(1–2), 287–317,
doi:10.1007/s10584-010-9856-z.
Vano, J. A., T. Das, and D. P. Lettenmaier (2012), Hydrologic sensitivities of Colorado River runoff to changes in precipitation and temperature, J. Hydrometeorol., 13, 932–949, doi:10.1175/JHM-D-11-069.1.
Vano, J. A., et al. (2014), Understanding uncertainties in future Colorado River streamflow, Bull. Am. Meteorol. Soc., 95, 59–78, doi:10.1175/
BAMS-D-12-00228.1.
Wenger, S. J., C. H. Luce, A. F. Hamlet, D. J. Isaak, and H. M. Neville (2010), Macroscale hydrologic modeling of ecologically relevant flow
metrics, Water Resour. Res., 46, W09513, doi:10.1029/2009WR008839.
Wood, A. W., R. L. Leung, V. Sridhar, and D. P. Lettenmaier (2004), Hydrologic implications of dynamical and statistical approaches to downscaling climate model outputs, Clim. Change, 62(1), 189–216, doi:10.1023/B:CLIM.0000013685.99609.9e.
VANO ET AL.
C 2015. American Geophysical Union. All Rights Reserved.
V
1976